Verdict Prediction using Artificial Neural Networks

Artificial Intelligence is making tremendous improvements in the judiciary system in various roles such as Verdict Prediction, Case Prioritization, and Recommendation Systems, among others. Various systems have been developed to automate the process of Verdict Prediction. Often, the judiciary may fail to deliver justice to common people due to corruption or other factors. This is one of the drawbacks in the modern judiciary system, where the verdict can be influenced by wealth. However, an automated system accessible to everyone can ensure a just verdict, thus increasing transparency. There are approximately 41,861,760 pending cases in the Supreme Court [16], and many of these cases can be resolved with fewer hearings. Consequently, a model has been constructed using neural networks to expedite and enhance accuracy. Through Natural Language Processing and deep learning, we have created a model that takes textual information about facts and evidence and provides the result of the winner and the appropriate punishment, along with a valid explanation. Utilizing NLTK libraries, the model has been trained on a vast amount of information about different cases using various neural networks, including CNN, RNN, and other simple neural networks. The stored model is employed for predicting the winner. Employing explanation-agnostic models, the sentences contributing to the result are displayed, and these sentences are subsequently sent to a model trained using the Indian Penal Code dataset to determine the punishment to be given to the culprit. This streamlined process optimizes court proceedings and ensures timely resolutions.